CES Briefing: Agency compensation models in the AI era, a speedrun of the CES show floor & Disney’s tech showcase

A change to how clients pay agencies seems inevitable in the AI era. How the agency compensation model should change, though, is anyone’s guess. But it has very much been a topic of discussion during CES this week.

“There has to be a different model,” said Raja Rajamannar, chief marketing and communications officer at Mastercard.

And agency executives are well aware that their compensation models need to change. Advertisers typically pay agencies based on the amount of time agencies spend working on brands’ businesses. But agencies are increasingly enlisting AI tools to do some of that work, like crunching numbers and compiling drafts of campaign reports, and the AI tools can complete those tasks in a fraction of the time it would take agency’s human employees.

“We are in the bill-by-the-hour business, and AI is a bill for either outcomes or a bill for platform or product. We’ve been trying to figure out what that model looks like,” said Rob Silver, evp and head of media at Razorfish.

“I have clients asking me that same question…. They’re asking, ‘If I’m paying for 100% of analyst and suddenly there’s a bot that’s doing 30% of their work, then where is my savings?” said Arthur Fullerton, global CTO of Havas CX.

So… where is the savings?

“We’re not at a point where we’re able to quantify the savings in time, but I anticipate that day is going to come. We’re going to have figure that out,” Fullerton said.

In fact, Havas is in the process of trying to figure that out. To start, the agency needs to quantify how much time is actually saved – or not – by effectively outsourcing tasks to AI tools. So it is setting up multivariate tests to compare how long it would take, say, a strategist and a data scientist to complete a task vs. if that work was handed off to an AI tool. But the comparison has to account for more than simply the time taken to complete a task. Other factors to consider are what any time savings enable when it comes to freeing up human employees to handle other work as well as how much time is needed for human employees to oversee and tweak the AI’s output.

“I don’t think there’s going to be many instances beyond automation tasks where we set it and forget it. But we do need to establish these measurement frameworks and determine how is this improving what we do,” Fullerton said.

But then agencies like Havas and Razorfish also need to determine how these improvements don’t come at a cost to their client billings. Some costs are easier to account for than others. 

An AI agent that an ad agency built to handle focus group research is a more tangible line item. “But if we then use it on the media management optimization side to write better briefs, to translate copy, to help us think through optimizations, that’s a little less clear as to how you would charge for that,” Silver said.

Mastercard’s Rajamannar laid out three potential models. One would be simply paying a flat fee on a per-project basis. Another would be the cost-plus model of paying for some combination of agency full-time employee hours plus the amount of tokens that would be consumed by AI tools (tokens are like gasoline for generative AI tools). And a third would be compensating agencies based on outcomes, such as a brand identifying it wants to drive a percentage lift in product sales and offering to pay the agency a given amount per each percentage point increase.

“I am willing to pay per point, say, $100,000. If you [as the agency] manage to get that increase [at a cost to the agency of] $5,000 and you pocket $95,000, God bless you. But I am looking at it from my perspective: what is each percentage worth to me?” Rajamannar said.

Sounds simple enough. Which means of course it’s not. “Outcome-based compensation models should be the future,” said Digitas CEO Amy Lanzi. But there’s a but. 

“The hard part is the client’s ability to actually truly measure this performance metric [i.e. the outcome on which the agency’s compensation hinges,” Lanzi said. “Most clients don’t control all the factors that would allow that 1% increase, so then you’re not going to bet your compensation if it’s dependent on supply chain or the retailer accepting a new whatever or the sales guy showing up. There’s a lot of variables that make it hard for us to agree to something.”

Womp womp. Oh well. Maybe ChatGPT can come up with the answer.

Speedrun of the CES show floor

The Las Vegas Convention Center is ostensibly the epicenter of CES. But not really. At least not when it comes to the media and marketing executives in attendance who hardly leave the alphabetical Bermuda Triangle of the Aria-Bellagio-Cosmopolitan. So for those executives unable to make it to the CES show floor – as well as anyone who steered clear of Sin City altogether – here’s a speedrun of what they (may have) missed.

Disney’s programmatic pitch goes live

Programmatic’s real-time bidding technology is finally fast enough for live sports – was, more or less, the gist of Disney’s Global Tech & Data Showcase event on Wednesday.

During the presentation, Disney executives talked up the company’s recent moves to equip its live sports inventory for programmatic buying – and vice versa. 

• A certification program for ad tech firms – starting with Google’s Display & Video 360, The Trade Desk and Yahoo DSP as well as Magnite – to programmatically access live sports and entertainment inventory and support real-time adjustments, like upping advertisers’ frequency caps to pounce on viewership spikes.
• Support for advertisers to deliver different ad creative to audiences from a single advertiser in the same ad slot during ESPN livestreams. 
• Keyword-based targeting for live programming to serve up ads connected to what was said during a live show, with a way to dynamically assemble the ad creative on the fly.

The announcements are the latest fruits born of Disney having its own ad server, which the company had gained through its acquisition of Hulu. 

Disney has been expanding the use of that ad server across its properties, which had previously run on Google’s ad server, and now it is extending the Disney Ad Server to ESPN, just in time for the launch of ESPN’s standalone streaming service later this year. Which was why so much of Disney’s presentation highlighted the intersection of programmatic and live sports.